Cg / Cgk (Gage Capability)
Gage capability vs tolerance — quick acceptance check for new measuring systems.
Serves these standards
About this tool
Cg and Cgk answer a narrower question than Cpk: is this measuring system good enough for the tolerance it has to check? The study measures a reference standard repeatedly (typically 50 readings) and compares the scatter of the readings with a fixed share of the tolerance — usually 20%.
Cg looks only at the spread (precision); Cgk additionally captures whether the mean of the readings sits where it should (bias). The customary acceptance benchmark is Cg ≥ 1.33 and Cgk ≥ 1.33, though many automotive customers tighten this to 2.0 for new gages.
Frequently asked questions
Cg vs Cgk — what is the difference?
Cg evaluates precision only: how tightly repeated readings cluster. Cgk adds location: whether the average reading is centred on the reference value. A gage can show an excellent Cg and still fail Cgk if it reads consistently high or low — that is bias, and it calls for adjustment or correction rather than replacement.
What is an acceptable Cgk value?
Cgk ≥ 1.33 is the common acceptance rule, equivalent to the gage consuming no more than 20% of the tolerance. For new gage acceptance, many automotive standards demand Cg/Cgk ≥ 2.0. Below the limit: correct bias first, then attack repeatability (fixture, cleaning, technique).
How are Cg and Cgk related to Cpk?
They are cousins, not the same thing. Cpk judges the production process against its tolerance using process variation; Cg/Cgk judge the measuring system against the tolerance using repeated readings of a reference standard. Do a Cg/Cgk study before trusting any Cpk or SPC data the gage produces.
How is a Cg study performed?
Measure a suitable reference standard (typically in the middle of the tolerance) 50 times under the same conditions, without operator interaction between readings. Compute the standard deviation of the readings and compare it with 20% of the tolerance; Cgk additionally accounts for the bias between the mean reading and the reference value. This tool runs the full calculation.